A Novel Deep CNN Model for COVID-19 Detection based on Lung CT Images

Abstract

In recent months, worldwide society has been destroyed by COVID-19 (Coronavirus Disease 2019), a catastrophic contagious illness that has ravaged the whole world. Deep learning (DL) approaches are the most widely used strategies for chest CT scan processing and illness prediction in the medical field. COVID-19 is infectious sickness that mostly influences the lungs also has the potential to be fatal if left untreated in its extreme form. It also affects the lung CT scans of individuals who are impacted. The use of lung CT scans has allowed us to develop a deep convolution neural network (DCNN)-based model that will make it easier to provide a more appropriate COVID-19 detection method throughout this pandemic. The dataset we used for training and testing consisted of 349 COVID-19 and 397 non-COVID-19 lung CT frames, which we collected for these reasons. In addition, the data augmentation approach was adopted, and 1744 CT frames of COVID-19 patients and 1588 CT frames of non-COVID-19 patients were obtained. Of these, we have utilized 80% of lung CT frames for training reasons and just 20% of the lung CT frames for testing purposes. Results of studies suggest that our models are successful under COVID-19 testing conditions. The test findings demonstrate that our method is superior. The experimented results reported that the proposed CNN algorithm attained overall accuracy of 99.79% which is higher than the existing model. With the findings of this study, it is possible to use DL to help in the objective of diagnosis of COVID-19 disease via CT scans of the lungs, which has a promising potential in terms of future research.